Measuring Employee Hours in Government Surveys
Prepared by the
Cognitive Applications for Establishment Surveys Working Group
Organizer: Clyde Tucker, Bureau of Labor Statistics
Sylvia Fisher, Bureau of Labor Statistics
Karen Goldenberg, Bureau of Labor Statistics
Eileen O’Brien, U.S. Bureau of the Census
Clyde Tucker, Bureau of Labor Statistics
Diane Willimack, U.S. Bureau of the Census
Presented to the Federal Economic Statistics Advisory Council
June 7, 2001
Measuring Employee Hours in Government Surveys
I. Introduction
A. Background
At the December, 2000 meeting of the Federal Economic Statistics Advisory Council, an
interagency team of behavioral scientists from the Bureau of Labor Statistics (BLS) and the U.S.
Census Bureau discussed the use of cognitive methods to improve establishment survey
questionnaires (Cognitive Interagency Working Group, 2000 and O’Brien, Fisher, and
Goldenberg, 2001). The paper prepared for the presentation included a review of cognitive
theory as it pertains to survey research, descriptions of several cognitive methods, and an
application of two of these methods in the evaluation of a questionnaire used in BLS’s Current
Employment Statistics survey program. Attendees at this session indicated that further
investigations of establishment surveys using cognitive methods would be desirable. In
particular, one suggestion was to evaluate how the same concept was operationalized in the
questions and instructions across several government surveys.
After considering several concepts, it was decided that the team would explore the
measurement of employee hours in both BLS and Census Bureau surveys. We identified the
following surveys that collect employee hours:
•
•
•
•
•
•
•
Current Employment Statistics (CES) Survey at BLS
Hours at Work Survey (HWS) at BLS
Occupational Safety and Health Survey (OSH) at BLS
Employer Cost Index Survey (ECI) at BLS
Annual Survey of Manufactures (ASM) at the Census Bureau (which is also part of the
Economic Census for manufacturing industries during the census years)
Economic Census of Mineral Industries (CMI) at the Census Bureau
Survey of Plant Capacity and Utilization (PCU) at the Census Bureau.
Although all of the surveys were analyzed, the focus of this paper will be on the CES,
ASM, and CMI. These are the surveys that publish estimates of employee hours. The HWS and
ECI will be discussed where appropriate.
The research was done in three stages. Once we identified the surveys, the team
conducted an expert review of the questionnaires. At the next stage, focus groups were held with
program managers, program staff involved in data collection, and users of the measures of
employment and employee hours. The final stage involved a comparison of the selected data
items from CES, ASM, and CMI.
Following this introduction, we briefly describe the surveys discussed in the paper. In
Section II we turn to the methods used across all stages of the research. Section III presents the
results of the analysis, and it includes a discussion of the prospects for measuring all employee
hours. Because comparisons between estimates are limited to manufacturing and mining, we do
not address measurement of nonsupervisory employees in the analysis. Section IV contains a
comparison of the actual estimates from the surveys. The final section draws conclusions and
asks several questions about future work in this area.
B. The Surveys
This section provides a brief background on the surveys covered in this analysis to help
put the results in context.
CES. The Current Employment Statistics survey is a monthly BLS survey with a sample
of approximately 370,000 business establishments, and is the source of data on over-the-month
change in U.S. payroll employment, hours, and earnings, by detailed industry. For purposes of
this research, the survey collects total employment, production (or nonsupervisory) worker
employment, and production (or nonsupervisory) employee hours. Panels of respondents report
monthly, some by mail or fax, some by Computer Assisted Telephone Interview (CATI). The
majority of respondents participate by Touchtone Data Entry (TDE). Regardless of mode, CES
respondents use questionnaires tailored to their establishments' industries to compile information.
This research examined the CES Manufacturing and Mining questionnaires, which are identical
except for the specifications of production workers.
HWS. The Hours at Work Survey is an annual BLS mail survey of about 6000
employers, with a sample drawn from the BLS ES-202 Longitudinal Database. The purpose of
the survey is to generate ratios of hours worked to hours paid, which feed into BLS productivity
analyses. Ratios are computed at the SIC major division (2-digit) level. The survey collects total
employment, which it benchmarks to the CES; production or nonsupervisory worker
employment; and total hours worked and hours paid for the production workers at the
establishment during the previous year. No employment or hours data based on the HWS are
published.
ASM. The Annual Survey of Manufactures (ASM) is a panel survey of establishments in
manufacturing industries conducted annually by the Census Bureau. It consists of a mailout/mail-back survey of approximately 55,000 selected establishments, supplemented by
administrative data for small employers and new businesses. The mail survey represents all
establishments that received a form in the previous economic census for manufacturing
industries. The ASM is conducted as part of the Economic Census during the census years,
which end in 2 and 7. At that time, comparable data are collected from all manufacturing
establishments above a specified size. For purposes of this research, the ASM (and the
quinquennial Economic Census for manufacturing industries) obtains and publishes information
on total employment and on production worker employment and hours.
CMI. The Census of Mineral Industries (CMI) is also part of the Economic Census, and
is collected by the Census Bureau every 5 years during years ending in 2 and 7. From a universe
of approximately 25,000 establishments, the mail component consists of approximately 15,000
of the large single-establishment companies and all of the multi-establishment companies with
payroll classified in the Mining Sector. Estimates for the remaining 10,000 establishments are
developed using industry averages in conjunction with data that are obtained from administrative
records of other Federal agencies. The CMI uses essentially the same questions and instructions
as the ASM (except for quarterly employment) to collect information on employment,
production worker employment, and production employee hours. Published CMI data include
total and production worker employment for the first quarter and annual production employee
hours.
ECI. The Employment Cost Index (ECI) is a quarterly index that measures the change
over time in the cost of labor, free from the influence of employment shifts among occupations
and industries. It measures that change for total compensation (wages, salaries, and employer
costs for employee benefits), for wages and salaries only, and for benefit costs only. The ECI is
collected from establishments in both the private sector and the Federal Government. Specific
job categories are selected to represent broad occupational categories within a participating
establishment. The ECI is being integrated into the National Compensation Survey (NCS)
beginning in 2001 and will be collected from a subset of the NCS sample. About 18,000 units in
the NCS sample have been designated as wage plus benefits schedules, and these units will
comprise the sample for the ECI. 1 Employee hours are collected for all the occupations covered
by the survey, not just for production occupations.
II. Methods
A. Expert Review
The preliminary step in evaluating the concepts under study was an expert review. In
general, expert review is a process by which experienced and knowledgeable content or methods
experts apply a systemic approach to evaluate a topic or content area. For the purposes of this
study, the team reviewed the surveys shown in Section I and assessed the way the concepts of
“all employees”(or "total employment"), “production workers,” and “employee hours,” have
been operationalized.
As we worked our way through the Expert Review, we identified three dimensions along
which to evaluate the different concepts. These are:
•
•
•
The survey reference period.
The strategy for asking questions on the survey instrument or data collection form.
The strategy for asking questions in the instructions for the survey.
The product of this activity is a summary of definitions and the differences and similarities
in operationalizing each of the study concepts (See the Appendix.). Information gained from the
expert review was used to generate a series of questions to be addressed with informed advisors
in group discussions.
B. Meetings with Advisors
Two meetings were conducted with expert advisors, who were generally program
managers or similar level representatives for each of the surveys under review. The first meeting
took place with two advisors from the Census Bureau, and represented three of their surveys: the
ASM, the CMI, and the PCU. A second meeting was conducted at the Bureau of Labor Statistics
with four advisors, representing the CES, the HWS, the OSH, and the ECI.
Study team members used questions identified during the expert review to query the
advisors. Questions covered the background and history of the survey, previous approaches to
1
ECI data collection procedures differ from those of the other surveys. The program does not use a fixed data
collection instrument or single set of questions, but rather depends on trained field economists to obtain data
meeting survey specifications from employer records. For this reason, we did not include the ECI in the expert
review. An ECI representative participated in the Survey Advisors meeting, but because of scheduling constraints
the program was not represented in the focus groups.
assessing the construct, problems in current operationalization(s), issues associated with
collecting data from respondents about each concept, and how the survey used each of the key
concepts. In addition, advisors were asked whether they had plans to consider changing the
current operationalization of the concepts or data collection procedures.
C. Focus Groups
Overview of Focus Group Methods. Focus groups are guided discussions conducted
with a small number of individuals selected because they are members of a target population. A
moderator guides a focus group discussion using a discussion guide. The purpose of a focus
group is to collect the opinions, attitudes, and beliefs of participants about a given topic or issue.
Focus Group Composition. Participants were recruited for three focus groups: one
group with staff from each agency involved in the day-to-day operations of collecting and
producing the data (the “producer focus group”), and two groups with users of the data ( “user
focus groups”). Eight agency staff members participated in the producer focus group and a total
of nine data users, primarily economists from BLS, the Bureau of Economic Analysis (BEA), the
Census Bureau, and the Federal Reserve Board (FRB), participated in the user focus groups.
Focus Group Protocol. A protocol was developed as a guide to structure each
discussion. The protocol contained specific questions that were intended to promote the flow of
ideas, encourage expression of participants' opinions, and address issues of particular importance
to the participants relative to the topics under study. Table 1 shows the topics addressed in the
producers’ focus group. Topics for the two user groups appear in Table 2.
TABLE 1: Topics for Producer Focus Group
FOR “TOTAL EMPLOYEES” AND “PRODUCTION WORKERS,” RESPECTIVELY:
•
•
•
•
•
•
What the concept represents in the data
collection process
What elements are included and excluded
How the concept is currently measured and
how it has been measured in the past
•
Problems respondents have providing data
for the concept
• Users’ opinions about the quality of data
collected for the concept
• Procedures necessary to collect data a
different way
FOR “HOURS WORKED” AND “HOURS PAID,” RESPECTIVELY
How the concept is currently collected and • Problems respondents have providing data
for the concept
how it has been measured in the past
•
Procedures necessary to collect data in a
Problems respondents have providing data
different way
for the concept
Users’ opinions about the quality of the
data collected for the concept
FOR “ALL EMPLOYEE HOURS”
•
•
Whether the survey has considered
collecting “all employee hours”
Benefits/limitations of collecting “all
employee hours”
•
Ways to operationalize and measure “all
employee hours”
•
Possible data users' interest in this concept
TABLE 2: Topics for User Focus Groups
•
•
•
•
•
•
•
•
•
FOR “TOTAL EMPLOYEES,” “PRODUCTION WORKERS,” AND “EMPLOYEE
HOURS,” RESPECTIVELY:
How data element is used
• What other users think about the data
What components the user assumes are
• Usefulness of the concept as currently
included within the concept
measured
Consumers/audiences that access this data
• Other issues
FOR “HOURS WORKED” AND “HOURS PAID,” RESPECTIVELY
• Other users’ opinions about each concept
How data element is used
• Does concept address user needs
How user defines each concept
• How can measurement be improved for
Comparison of concepts: similarities and
each concept
differences
• Other issues
Concept most useful to user and reasons
why this is so
FOR “ALL EMPLOYEE HOURS”
How users define “all employee hours”
• Benefits/limitations of collecting “all
employee hours”
Which data users would like to see “all
employee hours” collected and why
• How “all employee hours” could be
operationalized and measured
D. Estimates
Using the data. When we began this project, we thought it would be instructive to
compare estimates of data collected by BLS and the Census Bureau for the key concepts by
industry. We anticipated being able to match up what should be similar data items, and to see
the effect on survey estimates of any differences in conceptual bases. What we found,
unfortunately, was a more complex situation.
Not all surveys publish hours data. While all seven of the surveys we reviewed collect
data on employee hours, not all of them publish the data. Among BLS surveys, the CES
publishes average weekly hours for production workers in mining and manufacturing industries.
The HWS publishes ratios of hours worked to hours paid, figures that proved useful when we
compared BLS and Census Bureau hours data. The ECI publishes data only for selected
occupations. On the Census Bureau side, manufacturing production employee hours are
available from the ASM, and from economic censuses for years ending with 2 and 7. Mining
hours are available from the CMI for census years.
Industry. Economic statistics are generally obtained within and published by industry.
The late 1990s have been a period of transition for industrial classification, marked by the shift
from the Standard Industrial Classification (SIC) system to the North American Industry
Classification System (NAICS). The two systems have different conceptual structures, and one
consequence is that many of the industry categories are not comparable. The Census Bureau used
NAICS codes for the 1997 Economic Censuses. Since Census Bureau economic surveys follow
the census, they have published industry data on a NAICS basis since 1997. The CES will not
publish on a NAICS basis until 2003. As a consequence of the different NAICS implementation
schedules, we confined our analysis of estimates to 2-digit SIC-based data for manufacturing
from 1996. For Mining, only available from the Census Bureau in years of the economic
censuses (most recently 1997), we identified the NAICS industries that were available at the
detailed 3- or 4- digit level in CES publications from 1997, and compared them.
Presentation of data. There are agency differences in presentation as well as differences
in industrial classification structures. The most pertinent of these differences involves the
presentation of hours data. The Census Bureau publishes total aggregate hours for production
workers for the year, while the BLS CES survey publishes average weekly hours for production
workers. We addressed the problem of aggregate work hours versus average weekly hours by
converting both to a per worker annual figure. For Census, we divided the aggregate total hours
by the number of production workers. We annualized the BLS weekly hours data by multiplying
by 52, a decision that seemed reasonable given that BLS data include paid leave.
Comparing published estimates. We built a series of spreadsheets that allowed us to
compare total employment, production worker employment, and production worker hours
between the ASM and the CES and between the CMI and the CES. In the case of total
employment, we used Census Bureau data that include employment from auxiliary
establishments, because the CES data do not differentiate between auxiliary and other units. We
looked at the CMI for 1997 and compared it to the CES, but unfortunately there were only a few
NAICS mining industries that we could compare directly and that were also published in the
CES. In all cases, we computed ratios of published Census Bureau employment and production
worker figures to their published BLS counterparts. We also computed ratios of the normalized
hours estimates. As a further analytic tool, we compared the Census Bureau:CES hours ratio to
the hours worked:hours paid (HW:HP) ratios produced by the BLS HWS survey.
III.
Results
A.
The Concepts: Total Employment and Production Workers
Expert review. This research began with the intent to look at how the concept
"employee hours" has been operationalized by different agency surveys. Almost immediately,
however, it became clear that we could not look at hours without putting them into a broader
context. Most of the surveys collect hours for production workers (or nonsupervisory workers),
rather than for all employees. Therefore, we needed to look at how the surveys defined
production workers. We found that some survey instruments requested this count as a subset of
total employment, whereas others requested it as one of two components to be added to compute
total employment. These different question strategies change the context within which the
questions are asked. Therefore, we extended the expert review of employee hours to include the
concepts of total employment/all employees, production or nonsupervisory workers, and
production or nonsupervisory employee hours. We do not discuss results for nonsupervisory
workers in this paper.
Establishment surveys conducted by or for U.S. government agencies are directed in the
Statistical Policy Handbook to refer to "the payroll period containing the 12th of the month"
(OMB, 1978). Most of the surveys we reviewed conform to this guideline, but do so for
different points in the year. The ASM instructs respondents to include all employees reported to
the Internal Revenue Service on IRS form 941. Beginning in 2000, however, the survey has
asked employers to augment this figure with workers who might not be part of the 941 payroll
but who are part of the establishment's labor force--"co-employees" or "leased employees." The
CMI and the CES use payroll employment as their base for all employees. In addition, the CMI,
the CES, and the HWS specifically exclude employees of contractors, while the ASM does not
explicitly mention them.2
The team discovered that the Census Bureau and BLS surveys use different approaches to
obtain total employment and production worker employment on their data collection forms. The
CES questionnaire is a large matrix with a response row for each of 12 months. All Employees is
2
Professional employer organizations (PEOs) that take over administrative responsibilities for a staff and “lease” the
employees back to the establishment are a relatively new phenomenon. More research is needed, however, to
determine whether or not employers consider PEO employees to be contractors.
one of the column headings, followed by its definition, and Production Workers follow in
another column. HWS has questions with self-contained response spaces and built-in
instructions, including one describing how to report total employment. Both the CES and the
HWS use a "top down" strategy, in which they ask first for the total number of employees and
then for the number of production workers. The ASM and the CMI also ask for employment on
self-administered forms. These surveys follow a "bottom up" strategy, in which the survey first
requests the number of production workers and then the number for "all other employees." On
the ASM, the production worker component is the average number of production workers across
four quarters. The respondent adds this average to the number of All Other Employees to
compute total employment. Because of the bottom-up strategy, the ASM and the CMI forms
explicitly collect All Other Employees, a component of employment not shared by the BLS
surveys.
The team observed a variety of methods for presenting questions, definitions, and other
respondent materials. The CES is short and relatively self-contained, with instructions printed
on the back of the form. The HWS integrates the questions and instructions on the data collection
form. The ASM and the CMI have separate instruction booklets. Our review assumes that
respondents read and follow the instructions as they compile their data, although we know that
this frequently does not happen. The ASM and CMI instructions have headings called
Employment, which present global instructions describing who to count. In this sense the
instructions offer a top-down strategy, which is different from what appears on the survey
instrument.
All of the surveys discussed here have a similar sense of who is a production worker, but
they employ different methods to convey that information. Each survey has a set of detailed
"includes" and "excludes" for production workers. Some use bullet lists or short lists, while
others present lists in paragraph format. We know from research over the past fifteen years that
bullet lists are easier to read and understand than lists in paragraph format. We also see different
words, probably intended to mean the same thing: line-supervisor level, working supervisor
level, working supervisors who may be "in charge." The CMI defines production workers as
“PRODUCTION, DEVELOPMENT, AND EXPLORATION WORKERS.” This very specificsounding title offers no information about who or what might constitute the employees who are
to be counted here.
Advisors. Both the Census Bureau and BLS advisors noted that the core definitions for
the employment concepts have been extremely stable over time, but the population being
measured has undergone tremendous change. Any differences in the definitions in these surveys
over time primarily reflect these changes. The advisors from the Census Bureau were most
concerned with how well the current instrument designs capture the changing structure of
employment arrangements between worker and employer, e.g. contract labor, leased employees,
and temporary employees. Where ‘non-traditional’ employee arrangements are expanding or
taking over a plant, they worry that the Census Bureau may "lose" plants on their sampling
frames. There will be some attempt to measure the prevalence of these emerging employment
arrangements in the ASM 2001.
The advisors indicated that “production workers” is an important data element collected
on many BLS and Census surveys, and that its operationalization has been the same for many
years. This finding emerged even though the advisors acknowledged that there have been a
number of shifts in employment that might impact production worker data. Chief among these is
the increase in the number of workers who are not actual employees of the establishment, such as
leased employees and co-employees. It is still not clear whether the factory would have records
documenting leased employee hours or what they are being paid.
The CES advisor estimated that approximately 70 percent of employers now use standard
accounting or payroll software packages. Therefore, the CES (and other surveys) asks
respondents if their staff can talk to payroll or accounting personnel who report for the
establishment. The pervasive use of software packages may make it more difficult for
respondents to provide production worker data, since the payroll packages seldom, if ever, have
an option available to classify workers as “production workers.” Thus, respondents generally
have to make this distinction themselves in some other way in order to provide CES with
production worker estimates.
Producers. The first concern voiced by producers regarding data quality was
respondent difficulty in computing simple sums and averages. These types of error are readily
identified in automated edits whereas problems such as reporting data for the wrong pay periods
are more difficult to detect. Because of the relative infrequency of the Census economic
censuses and surveys, respondents are less familiar with the routine and logic of reporting
employment for the pay period including the 12th of March. It is also the case that 5-8% of CES
respondents are new to the task each month. Producers noted that the timing of internal payroll
reports was an impediment to firms reporting for the correct pay period in the CES. In the CES
and other surveys mentioned here, producers indicated that respondents often use available
records to estimate totals if information for the pay period of the twelfth is not available.
On the whole, this group of producers felt that establishments had little trouble reporting
either all employees or production workers. A larger issue, they believed, was reporting for a
designated sample establishment. Respondents sometimes have difficulty reporting for the
sampling unit defined for a survey. One example was mentioned here in connection with the
conduct of manufacturing censuses and surveys. The Census Bureau periodically sees
respondents having a problem assigning employee numbers where two distinct business
activities coexist at one location, e.g. a raw sugar operation (agriculture) and a sugar refinery
operation (manufacturing).
Several survey producers remarked they were concerned that respondents would not call
to obtain clarification about questions they were confused about. Another complication results
from the fact that many establishments have third parties, such as contracted accountants or
payroll companies, complete their government surveys. These third parties are unlikely to know
the establishment’s employee base well enough to be able to say who qualifies as a production
worker and who does not.
Users. Generally, users understood the way the all employee concept is defined in the
major surveys (CMI, ASM and CES). Users would like to include the self-employed, to get a
better measure of macroeconomic activity, to produce more complete productivity estimates and
so on. BEA, in particular, is interested in employment, hours, and earning measures for the selfemployed. Other users would like to have contract workers counted both where they are paid
and where they work in order to support better productivity estimation. Users and respondents
are both somewhat frustrated by the use of the pay period of the 12th as a reference. Whereas
respondents may be unaware of the logic using such a reference date, the users clearly would
prefer a more complete measure of total employment, perhaps a monthly measure, rather than a
sample of employment within the month. The concern with the current design is that it may miss
within-month variability and events such as strikes or shutdowns.
Whereas producers and advisors were using the all employee number for benchmarks, for
validity checks on sampled units, and as a key edit and frame variable, users were relying on this
figure as a key economic indicator. They use it, along with other elements, to measure economic
activity at the industry and national level, to understand the business cycle, and to support
macroeconomic models that forecast changes in the US economy. Because users favor
timeliness over completeness of the data, they rely more on the CES than on the more
comprehensive Census Bureau economic surveys and censuses.
A source of user frustration with the production worker data is that it is somewhat unclear
who is currently included in the production worker estimate, in particular, whether contract,
leased, and temporary employees are included. Users almost uniformly agree on a preferred
definition of production workers that can be effectively summarized by one user’s definition “as
nonsupervisory people on the factory floor who are not managers.” The preferred definition
includes leased and contract workers. The data are not always this comprehensive. Users report
that their optimal choice for production worker data would be to have number of workers, hours
they worked, and person day figures available for each plant, and, ideally, to have this data for a
month (preferably first to last of month).
B. The Concepts: Employee Hours
Expert review. Depending on the survey, employee hours means either hours worked or
hours paid, but, except for the ECI, hours are collected only for production workers. Hours
worked refers to all time spent on the job, including overtime, standby time, and time traveling
between job sites (but not commuting). Hours paid includes paid time worked as well as various
forms of paid leave. The Census Bureau surveys all collect hours worked. At BLS, the CES
collects hours paid, perhaps because it is benchmarked to the ES-202, which collects total wages
for hours paid. The HWS is a special-purpose survey that exists primarily to produce ratios of
hours worked to hours paid, and so collects both. The reference period for work hours for the
Census Bureau's ASM and CMI is the entire previous year. Among the BLS surveys, the
reference period for the HWS is the entire previous year, while for the CES it is always the pay
period including the 12th of a specific month.
Both question wording and question flow are issues for work hours items. The ASM asks
for "Plant hours worked by production workers (Annual)", and uses the word "Total" as part of a
subheading. The choice of "Plant hours" may be intended to convey the idea of work performed
in a factory setting. The CMI specifies "Hours worked by production, development, and
exploration workers in [year],” also with a subheading that begins with the words "Total annual
hours worked…." The different wording is reasonable given that development and exploration
work, especially, could take place in processing plants or settings other than existing mineral
extraction facilities. The CES asks for production employee hours paid, in a column headed
"Production Employee hours: Report the total production employee hours paid, including
overtime, for the pay period that includes the 12th of the month." The HWS strikes a similar
note: "In [year], how many hours, including overtime, were all production workers paid directly
from the employer? Annual Total, [year] (Include all production workers on the payroll at any
time during the year.)"
Question flow also is an aspect of the hours data elements as they appear on their
respective questionnaires. On the ASM and CMI, multi-part items dealing with payroll
immediately follow the questions on employment and production workers. Although one of the
subitems is production worker payroll, other subparts of the payroll item are not related to the
number of production workers and also lie between production worker employment and the
hours for those workers. The interruption could affect the way respondents seek out relevant
records. While the CES also has a payroll question following the count of production workers, it
refers specifically to production workers, so the context is the same. The HWS asks for
production employee hours paid immediately after the number of relevant employees. As noted
above, the HWS seeks both hours paid and hours worked. Since most employers have records
on hours paid, but not necessarily on hours worked, the questionnaire gives respondents the
option of reporting either hours worked or hours of paid leave.
The instructions reflect the question flow described above. Once again, instructions
provide considerably more information about what is to be reported on some of these
questionnaires than appears on the form itself. The CMI begins its directions by defining an
hour worked "as the work of one person for 1 hour," and encourages respondents to use records
if records are available. The direction has a bullet list of items to include, among which are "all
hours worked or paid (except hours paid for vacations, holidays, or sick leave)" and "actual hours
worked by an employee who elects to work during a vacation period." The expert review team
found the first of these items a bit confusing, because we thought that time paid but not worked
was paid leave, and that is specifically excluded. The ASM's directions are essentially the same
as those for the CMI, but in paragraph form, with an important exception. While the CMI
excludes contractor hours, the ASM tells respondents to include hours worked by co-employees.
Instructions on the BLS surveys for hours paid list various types of paid leave to include
as well as regular work hours. On the HWS, hours paid excludes unpaid leave, unpaid meal
time, and normal commuting time, while the direction for hours at work tells respondents to
exclude the hours actually used for paid holidays, vacation, sick leave, etc. On the CES, the
instruction tells respondents that hours paid is the sum of hours worked, hours paid for "portalto-portal, stand-by, or reporting time," and hours of paid leave. There are no “includes” or
“excludes” listed on the CES form.
Advisors. The meetings with survey advisors identified various complexities associated
with employee hours. In addition to the hours worked-hours paid distinction, additional issues
arose concerning hours worked and certain types of employee groups, such as piece workers.
The CES staff believes that it is possible to obtain an hours estimate for piece workers, even
though time is not the basis for their compensation. The HWS excludes piece workers and
commission workers, because of the difficulty respondents have with estimating the number of
hours worked or hours paid for these workers. On the Census Bureau side, the ASM excludes
paid leave, and advisors thought this was because the ASM is using BLS definitions for the
purposes of productivity measurement.
Producers. The data producers generally agreed that it was much easier for respondents
to provide an hours paid estimate than to report on employee hours worked. All of the producers
also reported that there were no significant planned upcoming changes in definitions or in ways
of collecting hours information. Some producers reported that there is a substantial amount of
missing hours data. This problem is more pronounced for nonsupervisory workers, but it also
exists in the case of production workers in the manufacturing sector. Establishments often
cannot provide this information, because they do not classify employees as production workers
but, instead, make distinctions on the basis of hourly vs. salaried or exempt vs. nonexempt.
Users. Users were asked questions about hours paid and hours worked data and their
preferences for different types of data in doing their economic analyses. The discussion
indicated that there were some common interests in specific data elements among economists in
different agencies, as well as some variation regarding what data users would like to have
available to them. Users were also quick to point out that economists are aware of the missing
information; indeed, many economists attempt to use proxy data to fill in gaps in existing data, or
at least, attempt to account for how missing hours data affect their economic models.
Users were asked to specify the hours data they used and how they used them. A
significant issue for these users was the difference between hours paid versus actual hours on the
job. Users ranged in their description of their uses of hours data. One user reported that “hours
are a cyclical indicator of the economy,” and that they “provide a better sense of overall demand
for labor.” Another uses hours data to assess long-run growth, which he described as a key part
of the economic forecasting process. A third user collects hours data from the CES to obtain a
“production workers hour measure of input to factories,” and also accesses production worker
hours data from the ASM and Census of Manufactures. He noted that although the ASM
exhibited some sampling error on employment, he found it very useful to compare the CES and
ASM production worker hours, because he believed they assess two very similar concepts.
When queried about their preferences for hours data, users almost uniformly reported
they would most like to have “hours worked” available. On the other hand, hours paid figures
are important to those studying employee earnings. A few economists would prefer to have
both data elements available to them, that is, both an hours worked and an hours paid estimate, as
the presence of both would give them the greatest amount of information. Further, an hours paid
estimate could be used to help assess the economic value of actual hours worked. One
economist reported he would like to have an hours worked figure, but would also like to have the
benefits and hours paid data; this user utilizes ECI data to obtain benefits measures. BEA uses
both measures, but its most critical need is for high quality estimates of compensation.
Users were asked to identify those components they believed should be included within
an hours worked figure. Several users agreed that telecommuting and actual time at the job site
should definitely be included. There was less agreement about commuting time to and from
work and other time “off the clock,” such as work at home after hours. The users recognized the
difficulty respondents could have estimating actual hours worked when at home (such as
accounting for frequent interruptions).
Users differed in their opinions about how to collect hours data for certain employee
groups where an hours worked figure is likely to be especially problematic, such as
pieceworkers. Some of the users were unsure how to collect “hours worked” versus “hours paid”
for these employees, while others stated they believed that hours data would still be meaningful,
even if only as a “proxy” indicator of worker productivity.
When asked what could be done to improve current data collection activities in the
surveys we studied, a number of economists reported they would like to obtain an “hours paid”
figure for supervisory workers, as well as an hours worked figure. One economist noted that
there is an approximately constant ratio between hours paid and hourly earnings; he was
particularly concerned that the incidence of “off the clock” work might be cyclical and not
reflected in hours paid data. As one economist pointed out, estimates of hours differ between the
CES and the Current Population Survey (CPS), with the CPS figure being higher. One type of
comparison made by some of the users was between the CES and CPS workweeks to determine
whether there was cyclical variability in the gap between these two estimates.
Users were concerned that the hours series is not benchmarked to any other data and
noted that movements in the data over time are especially important for this information. They
are concerned about the potential effect on time series of a substantive change in data collected,
despite the obvious potential gains to be made from such changes.
C. The Concepts: All Employee Hours
Advisors. We asked the advisors to consider the advisability of collecting a new item,
hours for all employees. They pointed out several issues that would have to be addressed,
including how the concept would be operationalized, whether it would be worth disrupting
existing time series to collect this new information, and whether users would be interested in
seeing this data.
The BLS is considering collecting hours paid for all employees, and may decide to do so
in the future for both the CES and the HWS. The CES staff is particularly concerned about
whether shifting data collection from production employee hours to all employee hours would
result in increased measurement error in respondent reporting of hours paid. The ECI usually
estimates hours for salaried workers, based on a standard workweek. An example of possible
impediments to ECI’s collection of all employee hours would be unpaid "mandatory overtime,"
which is unlikely to be readily collected from respondents. The ECI staff projects that a special
protocol would be needed to capture these and other variations in hours data.
When asked whether it would be better to ask for component parts of all employee hours
versus asking respondents for a total figure, the CES advisor indicated it would be necessary for
them to ask for distinct components. This would especially be true when there are different
payroll frequencies, for example, hourly versus salaried personnel could be paid on different
time schedules. Another important issue is that many user groups don't want to lose the
production workers series. The HWS staff is concerned primarily with following the CES
methodology, but they and would prefer to ask respondents to generate one total estimate. The
ECI staff doesn't want or need to collect all employee hours. Advisors expressed some
agreement that users might prefer that BLS switch from collecting hours paid to collecting hours
at work. Although there is a sense that government users would prefer all employee hours and
all employee payroll data, there is also a significant concern that changes to current data
collection activities would need to be minimal so as not to upset existing time series or disrupt
the agencies’ ability to meet user information needs. In any case, the costs and benefits of any
changes should be weighed carefully, especially if most users are concerned with trends and not
levels.
Regarding whether there are benefits and/or limitations to collecting all employee hours,
the ASM advisor believes that records are not available for salaried employees. There is no
record of whether ASM users have ever requested hours worked for all employees. The ASM
does not, in general, receive too many questions about hours items. Instead, they get queries that
focus on employment and payroll issues.
Users. Although data producers did not see much demand for all employee hours, users
generally agreed that they would like to have all employee hours available to them, if possible.
“All employee hours worked is the measure we want,” said one user, echoing the sentiments of
several of the participants. One of the most common drawbacks identified by users was the lack
of hours data for supervisors, even though all acknowledged that it would be difficult to collect
this information accurately.
Users derive needed figures in several ways. One uses information from the ECI benefit
series. Another pointed out that, for service industries, BEA assumes supervisory workers work
the same number of hours as nonsupervisory workers. Other users make the “leap” from blue
collar to white collar to measure the hours of non-production workers, but agree that they
sometimes miss some other worker groups such as janitorial or security workers.
Users recognized the difficulty of measuring hours worked by salaried workers but were
unable to provide much in the way of solutions. There was some discussion of moving to an
hourly/salaried concept, but some users were not sure this would solve the problem. Users
pointed out that data reported for salaried workers often were based upon a 40-hour-work-week
model, and agreed that respondents are likely to find it difficult to accurately estimate how many
hours salaried workers actually worked within a given week. Two users indicated that it was
more critical for them to have hours worked data for supervisors than a general estimate of hours
worked for all employees. Others would like to see monthly estimates of all employee hours,
because this periodicity would allow them to use all employee data in the construction of their
forecasting models and in observing ongoing economic trends. In any case, the users reported
they would prefer an hours worked figure; as one said, “we live with” the hours paid figure.
Most wanted to maintain the production worker series.
IV.
The Data
Cognitive methods are useful for pointing out differences in concept definitions and the
way these are operationalized on the survey forms, but it always pays to look at the data to see
what else might be uncovered. Table 3 provides comparisons between the CES and ASM for
total employment, numbers of production workers, and employee hours from 1996 in
manufacturing industries. 3 As noted in Section II, 1996 is the last year for which the Census
Bureau published economic data on an SIC basis. At that time, there was not a systematic effort
to collect information on “co-employees” on the ASM. The term “co-employee” was added to
the collection instrument in 2000.
Under SIC classifications, establishments can be either operating establishments or
auxiliaries. Auxiliary establishments are separate special-purpose establishments, such as
headquarters, warehouses, laboratories, and repair shops that provide support services for
operating establishments. Auxiliaries receive the same SIC classification as the operating
establishments they serve, but the ASM only collects survey data from the operating
establishments (manufacturing plants). In off-Census years such as 1996, the Census Bureau
obtains auxiliary unit employment from a separate source, the Company Organization Survey.
The CES makes no distinction between operating and auxiliary employment and gathers total
employment data from its sampled establishments according to industry (SIC).4 To make the
3
CES data for 1996 were downloaded from the BLS Internet site and reflect annual averages, not seasonally
adjusted, for that year. ASM data were taken from the PDF file of the 1996 Census Bureau publication, 1996
Annual Survey of Manufactures. M96(AS)-1. Statistics for Industry Groups and Industries, downloaded from the
Census Bureau Internet Site. Total employment comes from Table 1b.
4
Note that this changes under NAICS, because auxiliary units will have separate industry classifications that reflect
their true economic activities.
two surveys’ figures for total employment comparable, we included the employment in auxiliary
units in the ASM total employment numbers. It should be pointed out, however, that the Census
Bureau estimate for auxiliary workers in manufacturing was higher than the BLS estimate in
1996 (1.35 million vs. 0.8 million). We also computed the ratio of ASM to CES total
employment. Notice that in some cases the ASM is higher and in others the CES is higher. For
a majority of the industries, the two numbers are within 5 percent of each other. These
discrepancies could be a result of the different ways the surveys go about collecting total
employees as reported in Section III. Other possible explanations include sampling error and
differences in the classification of units by industry.
In some cases, particularly for transportation equipment, tobacco products, and leather
and leather products, the proportional differences in the two estimates are sizeable. The latter
two are the two smallest manufacturing industries, and this could account for the large
differences. This is not true for transportation equipment, where the difference might be a
function of a sensitivity to the method of data collection, that is, the “top-down” versus “bottomup” approaches. Again, it also could be the result of different classification schemes.
Parallel data are presented for production worker employment. The differences there are
larger. The CES figures are usually greater than the ones from the ASM. Although many of the
discrepancies are within 5 percent, they can range as high as 29 percent, and six of the industry
differences are greater than 10 percent. These differences could be the result of the way the
concept is operationalized in the two surveys, but could also be caused by the differences in the
classification of operational and auxiliary units as described above.
The results in Table 3 indicate that, as would be expected, the average hours paid per year
from the CES is larger than the average number of hours worked from the ASM. More
interesting, a comparison between the ratio of ASM to CES production worker hours and the
ratio of Hours Worked to Hours Paid from the HWS reveals that the former ratio is consistently
larger than the latter.5 The correlation between the two ratios is a modest 0.4. Of course, it is
possible that our efforts to make the two figures comparable were not successful, but it is just as
likely that the methods for measuring hours worked in the ASM and HWS are not comparable.
As we learned from our research, hours worked is more difficult to measure than hours paid.
5
Of course, the ASM and the CES were not designed to produce a ratio of hours worked to hours paid.
Table 4 gives the same information for comparing estimates from mining industries. In
this case, only comparisons at the NAICS code level for selected industries can be made, and we
were unable to obtain auxiliary unit information for these industries. Thus, the comparisons for
total employment suffer. For production worker employment, the BLS numbers again are
higher. There are discrepancies between the ratios of ASM to CES production worker hours and
Hours Worked to Hours Paid from HWS, but they are in the opposite direction from those in
manufacturing.
After analyzing these data, we recontacted our users from the focus groups. When asked
about the differences we found, most of them again reported that they only use one of the
surveys and not both. Some knew about the different treatment of auxiliaries in the two surveys,
and some did not. Since most of the analysts are looking at changes over time and not levels, the
levels differences are less important. They have observed about the same changes from both
surveys. Being able to benchmark to universe files also reduces the concern. On the other hand,
for those interested in knowing about the level of production worker employment, there are some
rather large differences between the two surveys. Interestingly, some were not aware of the
changes in the treatment of auxiliaries under NAICS, where they will be in a separate category as
in the current ASM.
Table 3. Comparing CES and ASM for 2-digit SIC groups, 1996
(ASM All Employees include both operating and auxiliary establishments)
CES AE ASM AE AE
CES PW ASM PW PW
CES PW CES PW
ASM PW
Annnual
(000s) Ratio Annual
(000s)
Ratio Average Average
Total
average
ASM: average
ASM: Weekly hours paid hours in
(000s)
CES
(000s)
CES
Hours
per year,
millions
Paid
52 week
year
Employment, not seas adjusted
Manufacturing
Durable goods
Lumber and wood products
SIC
2400
1996
1996
1996
1996
18495.0 18665.6 1.009 12776.0
10789.0 10814.1 1.002 7386.0
778.4
758.9 0.975
639.6
1996
1996
12167.8 0.952
7118.6 0.964
612.8 0.958
ASM PW
Average
hours
worked per
year
Ratio
ASM:
CES
BLS
Difference,
1996 (ASM:CES) HWS
(HW:HP)
Ratio
HW:HP
1996
41.6
42.4
40.8
1996
2163.200
2204.800
2121.600
1996
25010.2
14718.0
1248.6
1996
2055.441
2067.541
2037.533
0.950
0.938
0.960
0.918
0.915
0.949
0.032
0.023
0.011
Furniture and fixtures
Stone, clay, and glass products
Primary metal industries
Fabricated metal products
Industrial machinery and
Equipment and computer equipment
2500
3200
3300
3400
3500
504.3
543.8
710.5
1448.7
2114.6
527.0
549.6
710.7
1529.0
2112.8
1.045
1.011
1.000
1.055
0.999
398.4
423.1
553.4
1088.3
1320.9
414.2
402.3
542.9
1114.4
1287.0
1.040
0.951
0.981
1.024
0.974
39.4
43.3
44.2
42.4
43.1
2048.800
2251.600
2298.400
2204.800
2241.200
836.7
835.8
1170.9
2338.8
2685.1
2020.039
2077.554
2156.751
2098.708
2086.325
0.986
0.923
0.938
0.952
0.931
0.932
0.931
0.913
0.918
0.916
0.054
-0.008
0.025
0.034
0.015
Electronic and other electrical
Equipment
3600
1660.6
1724.0
1.038
1056.0
1006.3
0.953
41.5
2158.000
2030.8
2018.086
0.935
0.915
0.020
Transportation equipment
Instruments and related products
Miscellaneous manufacturing
Industries
3700
3800
3900
1784.9
855.4
387.8
1599.8
889.4
412.9
0.896
1.040
1.065
1209.6
423.1
273.3
1032.2
423.1
283.4
0.853
1.000
1.037
44.0
41.7
39.7
2288.000
2168.400
2064.400
2153.2
850.3
567.8
2086.030
2009.690
2003.529
0.912
0.927
0.971
0.891
0.890
0.928
0.021
0.037
0.043
2000
2100
2200
2300
2600
2700
2800
2900
3000
7706.0
1691.9
41.4
626.5
867.7
683.6
1540.3
1033.8
142.1
982.7
7851.5
1643.1
45.9
603.5
906.0
677.4
1614.6
1081.0
141.3
1057.2
1.019
0.971
1.109
0.963
1.044
0.991
1.048
1.046
0.994
1.076
5390.0
1253.7
32.0
529.4
711.2
519.0
841.3
575.4
92.0
762.0
5049.2
1112.6
22.7
489.0
726.5
487.5
800.7
476.8
69.1
799.7
0.937
0.887
0.709
0.924
1.022
0.939
0.952
0.829
0.751
1.049
40.5
41.0
40.0
40.6
37.0
43.3
38.2
43.2
43.6
41.5
2106.000
2132.000
2080.000
2111.200
1924.000
2251.600
1986.400
2246.400
2267.200
2158.000
10292.2
2300.8
45.7
1015.8
1356.3
1051.0
1573.4
1014.2
160.6
1649.5
2038.382
2067.949
2013.216
2077.301
1866.896
2155.897
1965.031
2127.097
2324.168
2062.648
0.968
0.970
0.968
0.984
0.970
0.957
0.989
0.947
1.025
0.956
0.922
0.927
0.894
0.937
0.945
0.903
0.919
0.903
0.914
0.919
0.046
0.043
0.074
0.047
0.025
0.054
0.070
0.044
0.111
0.037
3100
95.7
81.5
0.852
73.9
64.6
0.874
38.1
1981.200
124.9
1933.437
0.976
0.945
0.031
Nondurable goods, all employees
Food and kindred products
Tobacco products
Textile mill products
Apparel and other textile products
Paper and allied products
Printing and publishing
Chemicals and allied products
Petroleum and coal products
Rubber and miscellaneous plastics
Products
Leather and leather products
Table 4. Mining Industries, NAICS Based Data, 1997
Total Employment
AE
CES
Census of
AE
mineral (000s)
Industries
1997
(Mar)
NAICS97
211111
Ratio
PW
Census: Census of
BLS
mineral
Industries
Mar-97 1997.00
1997
Production
Production Employee hours
Workers
CES
PW
PW Hours
PW
CES CES Avg.
Ratio
PW
Ratio
Worked
Census
PW
hours paid Census:
Census: Census of Avg hours
Avg.
per year,
BLS
BLS
mineral worked per weekly 52 week
Industries
year
hours
year
paid
Mar-97 1997.00
1997
1997
Mar-97
1997.00
1997
SIC
Crude Petroleum and Natural Gas
Extraction
1311
Bituminous Coal Mining
1220
100308
143.1
0.70
86699
90.4
0.96
58289
81.9
0.71
116712
2002.299
156303
2098.900
42.6
2215.200
0.904
46.5
2418.000
0.868
74469
74.2
1.00
212111
Bituminous Coal and Lignite
Surface Mining
1221
36502
30339
-
-
64682
-
-
-
-
212112
Bituminous Coal Underground
Mining
1222
50197
44130
-
-
91621
-
-
-
-
0.97
15326
212210
Iron Ore Mining
1011
7920
8.5
0.93
6787
7.0
2258.141
48.6
2527.200
0.894
V. Discussion
All of the surveys discussed here have a similar sense of who is a production worker, but
they employ different methods to convey that information. The HWS includes the details as part
of the question, while the CES puts them on the back of a page. The ASM and the CMI are
longer, more detailed surveys, and place detailed guidelines in separate booklets--a practice
questioned by Dillman (2000) on several grounds, including the increased likelihood that the
instructions will be ignored. In some cases the survey instrument appears to approach a topic in a
way that is inconsistent with the detailed instructions.
Does it matter whether a respondent looks first at total employment and then at
production workers, or the reverse? The top-down and bottom-up approaches require very
different cognitive tasks. Implicit in both types of request is the expectation that employers can
and do identify production workers in their records, an expectation that often is not warranted
(Goldenberg, 1994; Goldenberg and Stewart, 1999). Therefore, the way the task is presented
could matter, if total employment is available in records, while groups approximating production
workers and "all other employees" are not. Or it might matter if a respondent compiles
production workers and all other employees separately rather than subtracting one of them from
the total, because the two subgroups might not sum to the total on payroll records.
There are differences in format across the surveys. Some use bullet lists or short lists,
while others present lists in paragraph format. We know from research that bullet lists are easier
to read and understand than lists in paragraph format. We see different words, probably intended
to mean the same thing: line-supervisor level, working supervisor level, working supervisors
who may be "in charge." There are differences in the underlying population: BLS surveys and
the CMI do not include any type of contract employee, while the ASM asks for the inclusion of
leased employees.
As for employee hours, the different data collection schedules dictate that the surveys
will arrive at somewhat different estimates. The CES is conducted monthly, and the ASM
collects quarterly production worker figures. The CMI is collected only once every five years.
The Census Bureau’s surveys collect hours worked. Perhaps this is a more difficult concept to
measure than hours paid, which is the focus of the CES. Although the HWS measures hours
worked, it usually does so by subtraction. The BLS surveys and the Census Bureau surveys start
with a different frame, and, as we have discovered, treat auxiliaries differently in published
estimates. Although we feel we were able to achieve some conceptual comparability in the
published figures for total employment, we are less certain about the figures for production
workers. The differences in the way the employee hours numbers are collected make it difficult
to produce comparable yearly averages.
It appears that users do not always know the details behind the numbers from these
surveys. They were not even aware of some of the surveys we discussed. Although BEA has
noted differences between the estimates from the CES and the ASM, not much formal work has
been done to compare results across the surveys. On the other hand, some users were not
concerned with the differences we found, all the way from definitions to the final estimates, as
long as comparable changes in series were observed. Nonetheless, we find the differences
between the figures for production worker employment and the two ratios for hours worked to
hours paid somewhat troubling.
Regarding hours data for all employees, users clearly would like this information. Some
data producers, however, seem not to be aware of this desire. This could be because of the
different uses of the data by the external versus the internal users. Both producers and users
seem to understand the difficulty of collecting these data and both were concerned that only the
usual hours or a standard workweek would be collected for supervisory workers. Users could
not offer good suggestions on how to overcome this problem. Most would want to maintain the
production worker series, but create an hours series for other employees.
We have a number of ideas for further research on this topic. The reconciliation of the
differences noted in tables 3 and 4 is at the top of the list. The sample designs for BLS surveys
are quite different from the designs for the Census Bureau surveys. The impact of these
differences on the estimates should be studied in more detail. A comparison of the classification
of a matched sample of units by industry and by the operational/auxiliary designation should be
undertaken. To investigate the effects of the different collection strategies, cognitive interviews
with establishments should be done using the different survey instruments. Obtaining the
respondent’s point of view would round out the picture of potential measurement error begun by
considering the perspectives of data users, program managers, and data producers. We
concentrated only on users in the Federal government, but users from industry, academia, and
interest groups should be polled. We note, however, that this research should be done before any
changes are made in survey instruments or data collection procedures. We believe that time
series should not be disturbed unless there is strong evidence of measurement error.
We close with the following questions that need to be answered:
• Is it important to operationalize economic concepts in exactly the same way across
different surveys?
• What are the tradeoffs between improving measurement and disrupting time series?
• How important is it to collect hours for all employees? Any suggestions for how to do
it?
• What additional research in this general area should be done?
VI.
References
Cognitive Applications for Establishment Surveys Interagency Working Group. 2000. "Using
Cognitive Methods to Improve Questionnaires for Establishment Surveys: A
Demonstration." Presented to the Federal Economic Statistics Advisory Council,
Washington, D.C., December.
Dillman, Don A. 2000. Mail and Internet Surveys: The Tailored Design Method. New York,
Wiley.
Goldenberg, Karen L. 1994. "Answering Questions, Questioning Answers: Evaluating Data
Quality in an Establishment Survey." Paper presented at the American Association for
Public Opinion Research, Danvers, Massachusetts. Appears in Proceedings of the Section on
Survey Research Methods, American Statistical Association.
Goldenberg, Karen L. and Jay Stewart. 1999, "Earnings Concepts and Data Availability from
the Current Employment Statistics Survey: Findings from Cognitive Interviews." Joint
Statistical Meetings, Baltimore, MD. Appears in Proceedings of the Section on Survey
Research Methods, American Statistical Association.
O'Brien, Eileen, Sylvia Fisher, and Karen Goldenberg. Forthcoming. "Application of Cognitive
Methods to an Establishment Survey: A Demonstration Using the Current Employment
Statistics Survey." Presented at American Association for Public Opinion Research,
Montreal, Quebec, May 2001.
Office of Management and Budget, formerly Office of Federal Statistical Policy and Standards,
1978, Statistical Policy Handbook, Directive No. 12, Standard Definition of Payroll Periods
for Employment Reports.
Appendix
Concept
Annual Survey of Manufactures
Census Bureau
Total Employment Derived figure
• Reference period - Production workers (PW): Pay period
including the 12th for one month of a
specified quarter of the previous year
- Other employees: pay period 12th of
March
Strategy: Identify components of
• Strategy for
employment and sum them
asking on form
- Form asks for number of PW in
designated pay periods for each
quarter
- Form directs respondent to sum
quarterly PW figures and divide by 4
for annual average
- Form asks for All Other Employees
for pay period including March 12
- Form directs respondent to sum PW
average and All Other Employees to
get total
Separate instruction booklet
• Strategy for
- Describes who to count as employee
asking in
and co-employee (leased employee)
instructions
- Instructs respondents to exclude
agricultural employees from specific
industrial activities, as well as
proprietors and partners
- Lists who to include as PW
(paragraph)
- Directs respondent to compute
average number of PWs
- Lists employee categories to include
as All Other Employees (paragraph;
no pay period specified), and directs
respondents to exclude proprietors or
partners
- Employees reported on IRS form 941
• Basis of figure
- Also "co-employees" or "leased
employees" who work on site for
professional employer organizations
(PEOs)
Census of Mineral Industries
Census Bureau
Derived figure
Pay period that includes March 12th of
the previous year
Strategy: Identify components of
employment and sum them
- Form asks for number of production,
development, and exploration
workers in pay period including
March 12
- Form asks for All Other Employees
- Form directs respondents to sum PW
and All Other Employees for total
Current Establishment Statistics
(Mining and Manufacturing)
BLS
Record-based figure
Monthly: pay period including the
12th
Strategy: Start with total
- Form asks for total employment
for pay period including 12th of
reference month
Hours at Work Survey
BLS
Record-based figure
Pay period that includes March 12th
of the previous year
Strategy: Start with total
- Form asks for number of paid
employees who worked or
received pay for pay period
including 12th of March
Separate instruction booklet
Instructions on back of form
- Lists who to include and who to
- Lists employees to include in total
exclude based on payroll status and
employment (corporate officials,
specific criteria (e.g., paid on per-ton
employees on paid leave, part
or per-car basis)
timers)
- Lists workers to include as
- Lists employees to exclude from
production, development, exploration
total employment (proprietors and
workers by work function (bullets),
partners, unpaid family workers,
and directs respondents to exclude
people not in pay status during
supervisor and contractor employees
reference period, outside
from PW
contractors)
- Lists workers by function to include
as All Other Employees (bullets), with
direction to exclude contractor
employees
Instructions incorporated into
question
- Lists employees to include and
exclude (similar to CES)
"Full and part-time employees on the
payroll of this establishment" during
pay period including March 12
Persons who worked or received
pay for any part of the March pay
period
Persons who worked or received
pay for any part of the reference
pay period
Concept
Annual Survey of Manufactures
Production Workers
• Reference period Pay period including 12th of March,
May, August, November (months
3,5,8,11) of the previous year
- Form asks for number of PW for each
• Strategy for
quarter
asking on form
- Form directs respondent to compute
average number of PW per year
Separate instruction booklet
• Strategy for
- Lists categories of workers to include
asking in
as PW based on work functions
instructions
(paragraph)
- Lists workers who are not PW as All
Other Employees with direction on
who to include
- Directs respondents to include
workers up through the linesupervisor level
- Directs respondents to exclude
proprietors and partners.
Production Worker Hours
• Reference period Annual, previous year
Census of Mineral Industries
Current Establishment Statistics
(Mining and Manufacturing)
Hours at Work Survey
th
Pay period that includes March 12th of
the previous year
Monthly: pay period including the
12th
Pay period that includes March 12 of
the previous year
- Form asks for number of production,
development, and exploration
workers, pay period including
March 12
Separate instruction booklet
- Lists workers by function to include
as production, development, and
exploration workers
- Directs respondents to include
workers up through the working
supervisor level
- Directs respondents to exclude
supervisory employees and
employees of contractors
- Form asks for number of
employees who are production
workers
- Form asks for number of employees
who are production workers
Instructions on back of form
- (Mining) Lists different
occupations to include as PW by
mining or extraction industry
- (Manufacturing) Lists one set of
occupations to include for all
manufacturing industries
- Directs respondents to include
working supervisors and group
leaders who may be "in charge" of
a group of employees, but whose
supervisory functions are only
incidental to their regular work
- Directs respondents to exclude
managers and employees
performing any of the
nonproduction occupations shown
Instructions incorporated into
questions:
- Defines PW in terms of what they are
not: "employees whose major
responsibility is not to supervise,
plan, or direct the work of others."
- Contains list of employees to include
that combines work functions and
method of payment ("hourly and
salaried;" "recordkeeping primarily
related to production")
- Has list of occupations to exclude
that is broader than initial PW
definition
- Directs respondents to exclude piece
workers, commission-only workers
Annual, previous year
Monthly: pay period including the
12th
Hours paid
Annual, previous year
• Hours paid or
hours worked
• Strategy for
asking on form
"Total plant hours worked by PW"
Hours worked
Plant hours worked by PW (Annual)
Hours worked by production,
development, and exploration workers
in reference year (annual)
• Strategy for
asking in
instructions
Separate instruction booklet
- Includes all PW hours worked or paid
for except paid leave
- Includes overtime
- Includes hours worked by coemployees from professional
employer organization (PEO) or
employee leasing company
Separate instruction booklet
- Defines an hour worked
- Includes overtime
- Excludes paid leave
- Excludes contractor hours
- Excludes hours of proprietors or
partners
- Total PW hours paid, including
overtime, for the pay period that
includes the 12th of the month
- Separate collection of PW
overtime hours for manufacturing
(subset of total hours)
Instructions on back of form
Hours Paid is the sum of:
- Hours worked, including overtime
- Hours paid for portal-to-portal,
standby, or reporting time
- Hours of paid leave
Defines overtime:
- Hours for which overtime
premiums were paid because
hours exceeded scheduled hours
Hours paid and Hours worked or Hours
of paid leave
- Hours paid: Hours including overtime
that PW were paid directly from the
employer
- Collects "Hours of paid leave actually
used" as alternate data item if that is
available instead of hours worked
Instructions incorporated into
questions
Hours Paid:
- Includes paid leave
- Bases hours for salaried PW on
normally scheduled work hours
- Excludes piece rate and commissiononly workers
Hours Worked:
- Hours Paid minus paid leave